/[pcre]/code/trunk/pcre_exec.c
ViewVC logotype

Diff of /code/trunk/pcre_exec.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 608 by ph10, Sun Jun 12 16:25:55 2011 UTC revision 721 by ph10, Fri Oct 7 15:51:39 2011 UTC
# Line 57  possible. There are also some static sup Line 57  possible. There are also some static sup
57  #undef min  #undef min
58  #undef max  #undef max
59    
60  /* Values for setting in md->match_function_type to indicate two special types  /* Values for setting in md->match_function_type to indicate two special types
61  of call to match(). We do it this way to save on using another stack variable,  of call to match(). We do it this way to save on using another stack variable,
62  as stack usage is to be discouraged. */  as stack usage is to be discouraged. */
63    
64  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
# Line 76  negative to avoid the external error cod Line 76  negative to avoid the external error cod
76  #define MATCH_ACCEPT       (-999)  #define MATCH_ACCEPT       (-999)
77  #define MATCH_COMMIT       (-998)  #define MATCH_COMMIT       (-998)
78  #define MATCH_KETRPOS      (-997)  #define MATCH_KETRPOS      (-997)
79  #define MATCH_PRUNE        (-996)  #define MATCH_ONCE         (-996)
80  #define MATCH_SKIP         (-995)  #define MATCH_PRUNE        (-995)
81  #define MATCH_SKIP_ARG     (-994)  #define MATCH_SKIP         (-994)
82  #define MATCH_THEN         (-993)  #define MATCH_SKIP_ARG     (-993)
83    #define MATCH_THEN         (-992)
84    
85  /* This is a convenience macro for code that occurs many times. */  /* This is a convenience macro for code that occurs many times. */
86    
# Line 137  while (length-- > 0) Line 138  while (length-- > 0)
138    
139  /* Normally, if a back reference hasn't been set, the length that is passed is  /* Normally, if a back reference hasn't been set, the length that is passed is
140  negative, so the match always fails. However, in JavaScript compatibility mode,  negative, so the match always fails. However, in JavaScript compatibility mode,
141  the length passed is zero. Note that in caseless UTF-8 mode, the number of  the length passed is zero. Note that in caseless UTF-8 mode, the number of
142  subject bytes matched may be different to the number of reference bytes.  subject bytes matched may be different to the number of reference bytes.
143    
144  Arguments:  Arguments:
# Line 184  if (caseless) Line 185  if (caseless)
185  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
186    if (md->utf8)    if (md->utf8)
187      {      {
188      /* Match characters up to the end of the reference. NOTE: the number of      /* Match characters up to the end of the reference. NOTE: the number of
189      bytes matched may differ, because there are some characters whose upper and      bytes matched may differ, because there are some characters whose upper and
190      lower case versions code as different numbers of bytes. For example, U+023A      lower case versions code as different numbers of bytes. For example, U+023A
191      (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);      (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);
192      a sequence of 3 of the former uses 6 bytes, as does a sequence of two of      a sequence of 3 of the former uses 6 bytes, as does a sequence of two of
193      the latter. It is important, therefore, to check the length along the      the latter. It is important, therefore, to check the length along the
194      reference, not along the subject (earlier code did this wrong). */      reference, not along the subject (earlier code did this wrong). */
195    
196      USPTR endptr = p + length;      USPTR endptr = p + length;
197      while (p < endptr)      while (p < endptr)
198        {        {
# Line 209  if (caseless) Line 210  if (caseless)
210    /* The same code works when not in UTF-8 mode and in UTF-8 mode when there    /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
211    is no UCP support. */    is no UCP support. */
212      {      {
213      if (eptr + length > md->end_subject) return -1;      if (eptr + length > md->end_subject) return -1;
214      while (length-- > 0)      while (length-- > 0)
215        { if (md->lcc[*p++] != md->lcc[*eptr++]) return -1; }        { if (md->lcc[*p++] != md->lcc[*eptr++]) return -1; }
216      }      }
217    }    }
218    
219  /* In the caseful case, we can just compare the bytes, whether or not we  /* In the caseful case, we can just compare the bytes, whether or not we
220  are in UTF-8 mode. */  are in UTF-8 mode. */
221    
222  else  else
223    {    {
224    if (eptr + length > md->end_subject) return -1;    if (eptr + length > md->end_subject) return -1;
225    while (length-- > 0) if (*p++ != *eptr++) return -1;    while (length-- > 0) if (*p++ != *eptr++) return -1;
226    }    }
227    
228  return eptr - eptr_start;  return eptr - eptr_start;
# Line 276  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM Line 277  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM
277         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
278         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
279         RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,         RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
280         RM61,  RM62, RM63, RM64 };         RM61,  RM62, RM63 };
281    
282  /* These versions of the macros use the stack, as normal. There are debugging  /* These versions of the macros use the stack, as normal. There are debugging
283  versions and production versions. Note that the "rw" argument of RMATCH isn't  versions and production versions. Note that the "rw" argument of RMATCH isn't
# Line 383  typedef struct heapframe { Line 384  typedef struct heapframe {
384    int Xprop_type;    int Xprop_type;
385    int Xprop_value;    int Xprop_value;
386    int Xprop_fail_result;    int Xprop_fail_result;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
387    int Xoclength;    int Xoclength;
388    uschar Xocchars[8];    uschar Xocchars[8];
389  #endif  #endif
# Line 477  Returns:       MATCH_MATCH if matched Line 475  Returns:       MATCH_MATCH if matched
475    
476  static int  static int
477  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,
478    const uschar *markptr, int offset_top, match_data *md, eptrblock *eptrb,    const uschar *markptr, int offset_top, match_data *md, eptrblock *eptrb,
479    unsigned int rdepth)    unsigned int rdepth)
480  {  {
481  /* These variables do not need to be preserved over recursion in this function,  /* These variables do not need to be preserved over recursion in this function,
# Line 550  HEAP_RECURSE: Line 548  HEAP_RECURSE:
548  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
549  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
550  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
 #define prop_category      frame->Xprop_category  
 #define prop_chartype      frame->Xprop_chartype  
 #define prop_script        frame->Xprop_script  
551  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
552  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
553  #endif  #endif
# Line 590  declarations can be cut out in a block. Line 585  declarations can be cut out in a block.
585  below are for variables that do not have to be preserved over a recursive call  below are for variables that do not have to be preserved over a recursive call
586  to RMATCH(). */  to RMATCH(). */
587    
588  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
589  const uschar *charptr;  const uschar *charptr;
590  #endif  #endif
591  const uschar *callpat;  const uschar *callpat;
592  const uschar *data;  const uschar *data;
593  const uschar *next;  const uschar *next;
594  USPTR         pp;  USPTR         pp;
595  const uschar *prev;  const uschar *prev;
596  USPTR         saved_eptr;  USPTR         saved_eptr;
597    
598  recursion_info new_recursive;  recursion_info new_recursive;
599    
600  BOOL cur_is_word;  BOOL cur_is_word;
601  BOOL condition;  BOOL condition;
602  BOOL prev_is_word;  BOOL prev_is_word;
603    
# Line 610  BOOL prev_is_word; Line 605  BOOL prev_is_word;
605  int prop_type;  int prop_type;
606  int prop_value;  int prop_value;
607  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
608  int oclength;  int oclength;
609  uschar occhars[8];  uschar occhars[8];
610  #endif  #endif
# Line 632  int stacksave[REC_STACK_SAVE_MAX]; Line 624  int stacksave[REC_STACK_SAVE_MAX];
624  eptrblock newptrb;  eptrblock newptrb;
625  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
626    
627  /* To save space on the stack and in the heap frame, I have doubled up on some  /* To save space on the stack and in the heap frame, I have doubled up on some
628  of the local variables that are used only in localised parts of the code, but  of the local variables that are used only in localised parts of the code, but
629  still need to be preserved over recursive calls of match(). These macros define  still need to be preserved over recursive calls of match(). These macros define
630  the alternative names that are used. */  the alternative names that are used. */
631    
632  #define allow_zero    cur_is_word  #define allow_zero    cur_is_word
# Line 680  if (md->match_call_count++ >= md->match_ Line 672  if (md->match_call_count++ >= md->match_
672  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
673    
674  /* At the start of a group with an unlimited repeat that may match an empty  /* At the start of a group with an unlimited repeat that may match an empty
675  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
676  done this way to save having to use another function argument, which would take  done this way to save having to use another function argument, which would take
677  up space on the stack. See also MATCH_CONDASSERT below.  up space on the stack. See also MATCH_CONDASSERT below.
678    
679  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
# Line 705  for (;;) Line 697  for (;;)
697    {    {
698    minimize = possessive = FALSE;    minimize = possessive = FALSE;
699    op = *ecode;    op = *ecode;
700    
701    switch(op)    switch(op)
702      {      {
703      case OP_MARK:      case OP_MARK:
# Line 783  for (;;) Line 775  for (;;)
775      md->start_match_ptr = ecode + 2;      md->start_match_ptr = ecode + 2;
776      RRETURN(MATCH_SKIP_ARG);      RRETURN(MATCH_SKIP_ARG);
777    
778      /* For THEN (and THEN_ARG) we pass back the address of the bracket or      /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
779      the alt that is at the start of the current branch. This makes it possible      the branch in which it occurs can be determined. Overload the start of
780      to skip back past alternatives that precede the THEN within the current      match pointer to do this. */
     branch. */  
781    
782      case OP_THEN:      case OP_THEN:
783      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
784        eptrb, RM54);        eptrb, RM54);
785      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
786      md->start_match_ptr = ecode - GET(ecode, 1);      md->start_match_ptr = ecode;
787      MRRETURN(MATCH_THEN);      MRRETURN(MATCH_THEN);
788    
789      case OP_THEN_ARG:      case OP_THEN_ARG:
790      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1+LINK_SIZE],      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top,
791        offset_top, md, eptrb, RM58);        md, eptrb, RM58);
792      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
793      md->start_match_ptr = ecode - GET(ecode, 1);      md->start_match_ptr = ecode;
794      md->mark = ecode + LINK_SIZE + 2;      md->mark = ecode + 2;
795      RRETURN(MATCH_THEN);      RRETURN(MATCH_THEN);
796    
797      /* Handle a capturing bracket, other than those that are possessive with an      /* Handle a capturing bracket, other than those that are possessive with an
# Line 808  for (;;) Line 799  for (;;)
799      subject position in the working slot at the top of the vector. We mustn't      subject position in the working slot at the top of the vector. We mustn't
800      change the current values of the data slot, because they may be set from a      change the current values of the data slot, because they may be set from a
801      previous iteration of this group, and be referred to by a reference inside      previous iteration of this group, and be referred to by a reference inside
802      the group. If we fail to match, we need to restore this value and also the      the group. A failure to match might occur after the group has succeeded,
803      values of the final offsets, in case they were set by a previous iteration      if something later on doesn't match. For this reason, we need to restore
804      of the same bracket.      the working value and also the values of the final offsets, in case they
805        were set by a previous iteration of the same bracket.
806    
807      If there isn't enough space in the offset vector, treat this as if it were      If there isn't enough space in the offset vector, treat this as if it were
808      a non-capturing bracket. Don't worry about setting the flag for the error      a non-capturing bracket. Don't worry about setting the flag for the error
# Line 820  for (;;) Line 812  for (;;)
812      case OP_SCBRA:      case OP_SCBRA:
813      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
814      offset = number << 1;      offset = number << 1;
815    
816  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
817      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
818      printf("subject=");      printf("subject=");
# Line 841  for (;;) Line 833  for (;;)
833    
834        for (;;)        for (;;)
835          {          {
836          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
837          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
838            eptrb, RM1);            eptrb, RM1);
839          if (rrc != MATCH_NOMATCH &&          if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
840              (rrc != MATCH_THEN || md->start_match_ptr != ecode))  
841            RRETURN(rrc);          /* If we backed up to a THEN, check whether it is within the current
842            branch by comparing the address of the THEN that is passed back with
843            the end of the branch. If it is within the current branch, and the
844            branch is one of two or more alternatives (it either starts or ends
845            with OP_ALT), we have reached the limit of THEN's action, so convert
846            the return code to NOMATCH, which will cause normal backtracking to
847            happen from now on. Otherwise, THEN is passed back to an outer
848            alternative. This implements Perl's treatment of parenthesized groups,
849            where a group not containing | does not affect the current alternative,
850            that is, (X) is NOT the same as (X|(*F)). */
851    
852            if (rrc == MATCH_THEN)
853              {
854              next = ecode + GET(ecode,1);
855              if (md->start_match_ptr < next &&
856                  (*ecode == OP_ALT || *next == OP_ALT))
857                rrc = MATCH_NOMATCH;
858              }
859    
860            /* Anything other than NOMATCH is passed back. */
861    
862            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
863          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
864          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
865          if (*ecode != OP_ALT) break;          if (*ecode != OP_ALT) break;
866          }          }
867    
868        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
869        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
870        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
871        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
872    
873        if (rrc != MATCH_THEN) md->mark = markptr;        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
874        RRETURN(MATCH_NOMATCH);  
875          if (md->mark == NULL) md->mark = markptr;
876          RRETURN(rrc);
877        }        }
878    
879      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
# Line 873  for (;;) Line 887  for (;;)
887      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
888      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
889    
890      /* Non-capturing bracket, except for possessive with unlimited repeat. Loop      /* Non-capturing or atomic group, except for possessive with unlimited
891      for all the alternatives. When we get to the final alternative within the      repeat. Loop for all the alternatives.
892      brackets, we would return the result of a recursive call to match()  
893      whatever happened. We can reduce stack usage by turning this into a tail      When we get to the final alternative within the brackets, we used to return
894      recursion, except in the case of a possibly empty group.*/      the result of a recursive call to match() whatever happened so it was
895        possible to reduce stack usage by turning this into a tail recursion,
896        except in the case of a possibly empty group. However, now that there is
897        the possiblity of (*THEN) occurring in the final alternative, this
898        optimization is no longer always possible.
899    
900        We can optimize if we know there are no (*THEN)s in the pattern; at present
901        this is the best that can be done.
902    
903        MATCH_ONCE is returned when the end of an atomic group is successfully
904        reached, but subsequent matching fails. It passes back up the tree (causing
905        captured values to be reset) until the original atomic group level is
906        reached. This is tested by comparing md->once_target with the start of the
907        group. At this point, the return is converted into MATCH_NOMATCH so that
908        previous backup points can be taken. */
909    
910        case OP_ONCE:
911      case OP_BRA:      case OP_BRA:
912      case OP_SBRA:      case OP_SBRA:
913      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
914    
915      for (;;)      for (;;)
916        {        {
917        if (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */        if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP;
918    
919          /* If this is not a possibly empty group, and there are no (*THEN)s in
920          the pattern, and this is the final alternative, optimize as described
921          above. */
922    
923          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
924          {          {
         if (op >= OP_SBRA)   /* Possibly empty group */  
           {  
           md->match_function_type = MATCH_CBEGROUP;  
           RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, eptrb,  
             RM48);  
           if (rrc == MATCH_NOMATCH) md->mark = markptr;  
           RRETURN(rrc);  
           }  
         /* Not a possibly empty group; use tail recursion */  
925          ecode += _pcre_OP_lengths[*ecode];          ecode += _pcre_OP_lengths[*ecode];
         DPRINTF(("bracket 0 tail recursion\n"));  
926          goto TAIL_RECURSE;          goto TAIL_RECURSE;
927          }          }
928    
929        /* For non-final alternatives, continue the loop for a NOMATCH result;        /* In all other cases, we have to make another call to match(). */
       otherwise return. */  
930    
931        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, eptrb,
       RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, eptrb,  
932          RM2);          RM2);
933        if (rrc != MATCH_NOMATCH &&  
934            (rrc != MATCH_THEN || md->start_match_ptr != ecode))        /* See comment in the code for capturing groups above about handling
935          THEN. */
936    
937          if (rrc == MATCH_THEN)
938            {
939            next = ecode + GET(ecode,1);
940            if (md->start_match_ptr < next &&
941                (*ecode == OP_ALT || *next == OP_ALT))
942              rrc = MATCH_NOMATCH;
943            }
944    
945          if (rrc != MATCH_NOMATCH)
946            {
947            if (rrc == MATCH_ONCE)
948              {
949              const uschar *scode = ecode;
950              if (*scode != OP_ONCE)           /* If not at start, find it */
951                {
952                while (*scode == OP_ALT) scode += GET(scode, 1);
953                scode -= GET(scode, 1);
954                }
955              if (md->once_target == scode) rrc = MATCH_NOMATCH;
956              }
957          RRETURN(rrc);          RRETURN(rrc);
958            }
959        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
960          if (*ecode != OP_ALT) break;
961        }        }
962      /* Control never reaches here. */  
963        if (md->mark == NULL) md->mark = markptr;
964        RRETURN(MATCH_NOMATCH);
965    
966      /* Handle possessive capturing brackets with an unlimited repeat. We come      /* Handle possessive capturing brackets with an unlimited repeat. We come
967      here from BRAZERO with allow_zero set TRUE. The offset_vector values are      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
968      handled similarly to the normal case above. However, the matching is      handled similarly to the normal case above. However, the matching is
969      different. The end of these brackets will always be OP_KETRPOS, which      different. The end of these brackets will always be OP_KETRPOS, which
970      returns MATCH_KETRPOS without going further in the pattern. By this means      returns MATCH_KETRPOS without going further in the pattern. By this means
971      we can handle the group by iteration rather than recursion, thereby      we can handle the group by iteration rather than recursion, thereby
972      reducing the amount of stack needed. */      reducing the amount of stack needed. */
973    
974      case OP_CBRAPOS:      case OP_CBRAPOS:
975      case OP_SCBRAPOS:      case OP_SCBRAPOS:
976      allow_zero = FALSE;      allow_zero = FALSE;
977    
978      POSSESSIVE_CAPTURE:      POSSESSIVE_CAPTURE:
979      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
980      offset = number << 1;      offset = number << 1;
# Line 939  for (;;) Line 989  for (;;)
989      if (offset < md->offset_max)      if (offset < md->offset_max)
990        {        {
991        matched_once = FALSE;        matched_once = FALSE;
992        code_offset = ecode - md->start_code;        code_offset = ecode - md->start_code;
993    
994        save_offset1 = md->offset_vector[offset];        save_offset1 = md->offset_vector[offset];
995        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
# Line 947  for (;;) Line 997  for (;;)
997        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
998    
999        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1000    
1001        /* Each time round the loop, save the current subject position for use        /* Each time round the loop, save the current subject position for use
1002        when the group matches. For MATCH_MATCH, the group has matched, so we        when the group matches. For MATCH_MATCH, the group has matched, so we
1003        restart it with a new subject starting position, remembering that we had        restart it with a new subject starting position, remembering that we had
1004        at least one match. For MATCH_NOMATCH, carry on with the alternatives, as        at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1005        usual. If we haven't matched any alternatives in any iteration, check to        usual. If we haven't matched any alternatives in any iteration, check to
1006        see if a previous iteration matched. If so, the group has matched;        see if a previous iteration matched. If so, the group has matched;
1007        continue from afterwards. Otherwise it has failed; restore the previous        continue from afterwards. Otherwise it has failed; restore the previous
1008        capture values before returning NOMATCH. */        capture values before returning NOMATCH. */
1009    
1010        for (;;)        for (;;)
1011          {          {
1012          md->offset_vector[md->offset_end - number] =          md->offset_vector[md->offset_end - number] =
1013            (int)(eptr - md->start_subject);            (int)(eptr - md->start_subject);
1014          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1015          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
1016            eptrb, RM63);            eptrb, RM63);
1017          if (rrc == MATCH_KETRPOS)          if (rrc == MATCH_KETRPOS)
1018            {            {
1019            offset_top = md->end_offset_top;            offset_top = md->end_offset_top;
1020            eptr = md->end_match_ptr;            eptr = md->end_match_ptr;
1021            ecode = md->start_code + code_offset;            ecode = md->start_code + code_offset;
1022            save_capture_last = md->capture_last;            save_capture_last = md->capture_last;
1023            matched_once = TRUE;            matched_once = TRUE;
1024            continue;            continue;
1025              }
1026    
1027            /* See comment in the code for capturing groups above about handling
1028            THEN. */
1029    
1030            if (rrc == MATCH_THEN)
1031              {
1032              next = ecode + GET(ecode,1);
1033              if (md->start_match_ptr < next &&
1034                  (*ecode == OP_ALT || *next == OP_ALT))
1035                rrc = MATCH_NOMATCH;
1036            }            }
1037          if (rrc != MATCH_NOMATCH &&  
1038              (rrc != MATCH_THEN || md->start_match_ptr != ecode))          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
           RRETURN(rrc);  
1039          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
1040          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
1041          if (*ecode != OP_ALT) break;          if (*ecode != OP_ALT) break;
1042          }          }
1043    
1044        if (!matched_once)        if (!matched_once)
1045          {          {
1046          md->offset_vector[offset] = save_offset1;          md->offset_vector[offset] = save_offset1;
1047          md->offset_vector[offset+1] = save_offset2;          md->offset_vector[offset+1] = save_offset2;
1048          md->offset_vector[md->offset_end - number] = save_offset3;          md->offset_vector[md->offset_end - number] = save_offset3;
1049          }          }
1050    
1051        if (rrc != MATCH_THEN) md->mark = markptr;        if (md->mark == NULL) md->mark = markptr;
1052        if (allow_zero || matched_once)        if (allow_zero || matched_once)
1053          {          {
1054          ecode += 1 + LINK_SIZE;          ecode += 1 + LINK_SIZE;
1055          break;          break;
1056          }          }
1057    
1058        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1059        }        }
1060    
1061      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1062      as a non-capturing bracket. */      as a non-capturing bracket. */
1063    
# Line 1009  for (;;) Line 1069  for (;;)
1069      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1070      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1071    
1072      /* Non-capturing possessive bracket with unlimited repeat. We come here      /* Non-capturing possessive bracket with unlimited repeat. We come here
1073      from BRAZERO with allow_zero = TRUE. The code is similar to the above,      from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1074      without the capturing complication. It is written out separately for speed      without the capturing complication. It is written out separately for speed
1075      and cleanliness. */      and cleanliness. */
1076    
1077      case OP_BRAPOS:      case OP_BRAPOS:
1078      case OP_SBRAPOS:      case OP_SBRAPOS:
1079      allow_zero = FALSE;      allow_zero = FALSE;
1080    
1081      POSSESSIVE_NON_CAPTURE:      POSSESSIVE_NON_CAPTURE:
1082      matched_once = FALSE;      matched_once = FALSE;
1083      code_offset = ecode - md->start_code;      code_offset = ecode - md->start_code;
1084    
1085      for (;;)      for (;;)
1086        {        {
1087        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1088        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
1089          eptrb, RM64);          eptrb, RM48);
1090        if (rrc == MATCH_KETRPOS)        if (rrc == MATCH_KETRPOS)
1091          {          {
1092            offset_top = md->end_offset_top;
1093          eptr = md->end_match_ptr;          eptr = md->end_match_ptr;
1094          ecode = md->start_code + code_offset;          ecode = md->start_code + code_offset;
1095          matched_once = TRUE;          matched_once = TRUE;
1096          continue;          continue;
1097            }
1098    
1099          /* See comment in the code for capturing groups above about handling
1100          THEN. */
1101    
1102          if (rrc == MATCH_THEN)
1103            {
1104            next = ecode + GET(ecode,1);
1105            if (md->start_match_ptr < next &&
1106                (*ecode == OP_ALT || *next == OP_ALT))
1107              rrc = MATCH_NOMATCH;
1108          }          }
1109        if (rrc != MATCH_NOMATCH &&  
1110            (rrc != MATCH_THEN || md->start_match_ptr != ecode))        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
         RRETURN(rrc);  
1111        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1112        if (*ecode != OP_ALT) break;        if (*ecode != OP_ALT) break;
1113        }        }
1114    
1115      if (matched_once || allow_zero)      if (matched_once || allow_zero)
1116        {        {
1117        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1118        break;        break;
1119        }        }
1120      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
1121    
1122      /* Control never reaches here. */      /* Control never reaches here. */
# Line 1053  for (;;) Line 1124  for (;;)
1124      /* Conditional group: compilation checked that there are no more than      /* Conditional group: compilation checked that there are no more than
1125      two branches. If the condition is false, skipping the first branch takes us      two branches. If the condition is false, skipping the first branch takes us
1126      past the end if there is only one branch, but that's OK because that is      past the end if there is only one branch, but that's OK because that is
1127      exactly what going to the ket would do. As there is only one branch to be      exactly what going to the ket would do. */
     obeyed, we can use tail recursion to avoid using another stack frame. */  
1128    
1129      case OP_COND:      case OP_COND:
1130      case OP_SCOND:      case OP_SCOND:
# Line 1068  for (;;) Line 1138  for (;;)
1138        if (pcre_callout != NULL)        if (pcre_callout != NULL)
1139          {          {
1140          pcre_callout_block cb;          pcre_callout_block cb;
1141          cb.version          = 1;   /* Version 1 of the callout block */          cb.version          = 2;   /* Version 1 of the callout block */
1142          cb.callout_number   = ecode[LINK_SIZE+2];          cb.callout_number   = ecode[LINK_SIZE+2];
1143          cb.offset_vector    = md->offset_vector;          cb.offset_vector    = md->offset_vector;
1144          cb.subject          = (PCRE_SPTR)md->start_subject;          cb.subject          = (PCRE_SPTR)md->start_subject;
# Line 1080  for (;;) Line 1150  for (;;)
1150          cb.capture_top      = offset_top/2;          cb.capture_top      = offset_top/2;
1151          cb.capture_last     = md->capture_last;          cb.capture_last     = md->capture_last;
1152          cb.callout_data     = md->callout_data;          cb.callout_data     = md->callout_data;
1153            cb.mark             = markptr;
1154          if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);          if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1155          if (rrc < 0) RRETURN(rrc);          if (rrc < 0) RRETURN(rrc);
1156          }          }
# Line 1238  for (;;) Line 1309  for (;;)
1309    
1310      else      else
1311        {        {
1312        md->match_function_type = MATCH_CONDASSERT;        md->match_function_type = MATCH_CONDASSERT;
1313        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1314        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1315          {          {
1316            if (md->end_offset_top > offset_top)
1317              offset_top = md->end_offset_top;  /* Captures may have happened */
1318          condition = TRUE;          condition = TRUE;
1319          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1320          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1321          }          }
1322        else if (rrc != MATCH_NOMATCH &&  
1323                (rrc != MATCH_THEN || md->start_match_ptr != ecode))        /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1324          assertion; it is therefore treated as NOMATCH. */
1325    
1326          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1327          {          {
1328          RRETURN(rrc);         /* Need braces because of following else */          RRETURN(rrc);         /* Need braces because of following else */
1329          }          }
# Line 1258  for (;;) Line 1334  for (;;)
1334          }          }
1335        }        }
1336    
1337      /* We are now at the branch that is to be obeyed. As there is only one,      /* We are now at the branch that is to be obeyed. As there is only one, can
1338      we can use tail recursion to avoid using another stack frame, except when      use tail recursion to avoid using another stack frame, except when there is
1339      we have an unlimited repeat of a possibly empty group. If the second      unlimited repeat of a possibly empty group. In the latter case, a recursive
1340      alternative doesn't exist, we can just plough on. */      call to match() is always required, unless the second alternative doesn't
1341        exist, in which case we can just plough on. Note that, for compatibility
1342        with Perl, the | in a conditional group is NOT treated as creating two
1343        alternatives. If a THEN is encountered in the branch, it propagates out to
1344        the enclosing alternative (unless nested in a deeper set of alternatives,
1345        of course). */
1346    
1347      if (condition || *ecode == OP_ALT)      if (condition || *ecode == OP_ALT)
1348        {        {
1349        ecode += 1 + LINK_SIZE;        if (op != OP_SCOND)
       if (op == OP_SCOND)        /* Possibly empty group */  
1350          {          {
1351          md->match_function_type = MATCH_CBEGROUP;          ecode += 1 + LINK_SIZE;
1352          RMATCH(eptr, ecode, offset_top, md, eptrb, RM49);          goto TAIL_RECURSE;
         RRETURN(rrc);  
1353          }          }
1354        else goto TAIL_RECURSE;  
1355          md->match_function_type = MATCH_CBEGROUP;
1356          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1357          RRETURN(rrc);
1358        }        }
1359      else                         /* Condition false & no alternative */  
1360         /* Condition false & no alternative; continue after the group. */
1361    
1362        else
1363        {        {
1364        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1365        }        }
# Line 1305  for (;;) Line 1390  for (;;)
1390      break;      break;
1391    
1392    
1393      /* End of the pattern, either real or forced. If we are in a recursion, we      /* End of the pattern, either real or forced. */
     should restore the offsets appropriately, and if it's a top-level  
     recursion, continue from after the call. */  
1394    
     case OP_ACCEPT:  
1395      case OP_END:      case OP_END:
1396      if (md->recursive != NULL)      case OP_ACCEPT:
1397        {      case OP_ASSERT_ACCEPT:
       recursion_info *rec = md->recursive;  
       md->recursive = rec->prevrec;  
       memmove(md->offset_vector, rec->offset_save,  
         rec->saved_max * sizeof(int));  
       offset_top = rec->save_offset_top;  
       if (rec->group_num == 0)  
         {  
         ecode = rec->after_call;  
         break;  
         }  
       }  
   
     /* Otherwise, if we have matched an empty string, fail if PCRE_NOTEMPTY is  
     set, or if PCRE_NOTEMPTY_ATSTART is set and we have matched at the start of  
     the subject. In both cases, backtracking will then try other alternatives,  
     if any. */  
1398    
1399      else if (eptr == mstart &&      /* If we have matched an empty string, fail if not in an assertion and not
1400          (md->notempty ||      in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1401            (md->notempty_atstart &&      is set and we have matched at the start of the subject. In both cases,
1402              mstart == md->start_subject + md->start_offset)))      backtracking will then try other alternatives, if any. */
1403    
1404        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1405             md->recursive == NULL &&
1406             (md->notempty ||
1407               (md->notempty_atstart &&
1408                 mstart == md->start_subject + md->start_offset)))
1409        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1410    
1411      /* Otherwise, we have a match. */      /* Otherwise, we have a match. */
1412    
1413      md->end_match_ptr = eptr;           /* Record where we ended */      md->end_match_ptr = eptr;           /* Record where we ended */
1414      md->end_offset_top = offset_top;    /* and how many extracts were taken */      md->end_offset_top = offset_top;    /* and how many extracts were taken */
1415      md->start_match_ptr = mstart;       /* and the start (\K can modify) */      md->start_match_ptr = mstart;       /* and the start (\K can modify) */
# Line 1352  for (;;) Line 1424  for (;;)
1424      matching won't pass the KET for an assertion. If any one branch matches,      matching won't pass the KET for an assertion. If any one branch matches,
1425      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1426      start of each branch to move the current point backwards, so the code at      start of each branch to move the current point backwards, so the code at
1427      this level is identical to the lookahead case. When the assertion is part      this level is identical to the lookahead case. When the assertion is part
1428      of a condition, we want to return immediately afterwards. The caller of      of a condition, we want to return immediately afterwards. The caller of
1429      this incarnation of the match() function will have set MATCH_CONDASSERT in      this incarnation of the match() function will have set MATCH_CONDASSERT in
1430      md->match_function type, and one of these opcodes will be the first opcode      md->match_function type, and one of these opcodes will be the first opcode
1431      that is processed. We use a local variable that is preserved over calls to      that is processed. We use a local variable that is preserved over calls to
1432      match() to remember this case. */      match() to remember this case. */
1433    
1434      case OP_ASSERT:      case OP_ASSERT:
# Line 1366  for (;;) Line 1438  for (;;)
1438        condassert = TRUE;        condassert = TRUE;
1439        md->match_function_type = 0;        md->match_function_type = 0;
1440        }        }
1441      else condassert = FALSE;      else condassert = FALSE;
1442    
1443      do      do
1444        {        {
1445        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1446        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1447          {          {
1448          mstart = md->start_match_ptr;   /* In case \K reset it */          mstart = md->start_match_ptr;   /* In case \K reset it */
1449            markptr = md->mark;
1450          break;          break;
1451          }          }
1452        if (rrc != MATCH_NOMATCH &&  
1453            (rrc != MATCH_THEN || md->start_match_ptr != ecode))        /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1454          RRETURN(rrc);        as NOMATCH. */
1455    
1456          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1457        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1458        }        }
1459      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1460    
1461      if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);      if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);
1462    
1463      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
# Line 1408  for (;;) Line 1483  for (;;)
1483        condassert = TRUE;        condassert = TRUE;
1484        md->match_function_type = 0;        md->match_function_type = 0;
1485        }        }
1486      else condassert = FALSE;      else condassert = FALSE;
1487    
1488      do      do
1489        {        {
# Line 1419  for (;;) Line 1494  for (;;)
1494          do ecode += GET(ecode,1); while (*ecode == OP_ALT);          do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1495          break;          break;
1496          }          }
1497        if (rrc != MATCH_NOMATCH &&  
1498            (rrc != MATCH_THEN || md->start_match_ptr != ecode))        /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1499          RRETURN(rrc);        as NOMATCH. */
1500    
1501          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1502        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1503        }        }
1504      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1505    
1506      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1507    
1508      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1509      continue;      continue;
1510    
# Line 1472  for (;;) Line 1549  for (;;)
1549      if (pcre_callout != NULL)      if (pcre_callout != NULL)
1550        {        {
1551        pcre_callout_block cb;        pcre_callout_block cb;
1552        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1553        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1554        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1555        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
# Line 1484  for (;;) Line 1561  for (;;)
1561        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1562        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1563        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1564          cb.mark             = markptr;
1565        if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);        if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1566        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1567        }        }
# Line 1494  for (;;) Line 1572  for (;;)
1572      offset data is the offset to the starting bracket from the start of the      offset data is the offset to the starting bracket from the start of the
1573      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1574    
1575      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1576      save their starting points and reinstate them after the recursion. However,      re-instated afterwards. We don't know how many are started and not yet
1577      we don't know how many such there are (offset_top records the completed      finished (offset_top records the completed total) so we just have to save
1578      total) so we just have to save all the potential data. There may be up to      all the potential data. There may be up to 65535 such values, which is too
1579      65535 such values, which is too large to put on the stack, but using malloc      large to put on the stack, but using malloc for small numbers seems
1580      for small numbers seems expensive. As a compromise, the stack is used when      expensive. As a compromise, the stack is used when there are no more than
1581      there are no more than REC_STACK_SAVE_MAX values to store; otherwise malloc      REC_STACK_SAVE_MAX values to store; otherwise malloc is used.
     is used. A problem is what to do if the malloc fails ... there is no way of  
     returning to the top level with an error. Save the top REC_STACK_SAVE_MAX  
     values on the stack, and accept that the rest may be wrong.  
1582    
1583      There are also other values that have to be saved. We use a chained      There are also other values that have to be saved. We use a chained
1584      sequence of blocks that actually live on the stack. Thanks to Robin Houston      sequence of blocks that actually live on the stack. Thanks to Robin Houston
1585      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1586        a lot, so he is not to blame for the current way it works. */
1587    
1588      case OP_RECURSE:      case OP_RECURSE:
1589        {        {
1590          recursion_info *ri;
1591          int recno;
1592    
1593        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1594        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1595          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1596    
1597          /* Check for repeating a recursion without advancing the subject pointer.
1598          This should catch convoluted mutual recursions. (Some simple cases are
1599          caught at compile time.) */
1600    
1601          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1602            if (recno == ri->group_num && eptr == ri->subject_position)
1603              RRETURN(PCRE_ERROR_RECURSELOOP);
1604    
1605        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1606    
1607          new_recursive.group_num = recno;
1608          new_recursive.subject_position = eptr;
1609        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1610        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1611    
1612        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1613    
1614        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1615    
1616        /* Now save the offset data. */        /* Now save the offset data */
1617    
1618        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1619        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 1536  for (;;) Line 1624  for (;;)
1624            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));
1625          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1626          }          }
   
1627        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1628              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
1629        new_recursive.save_offset_top = offset_top;  
1630          /* OK, now we can do the recursion. After processing each alternative,
1631        /* OK, now we can do the recursion. For each top-level alternative we        restore the offset data. If there were nested recursions, md->recursive
1632        restore the offset and recursion data. */        might be changed, so reset it before looping. */
1633    
1634        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1635        cbegroup = (*callpat >= OP_SBRA);        cbegroup = (*callpat >= OP_SBRA);
# Line 1551  for (;;) Line 1638  for (;;)
1638          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1639          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
1640            md, eptrb, RM6);            md, eptrb, RM6);
1641            memcpy(md->offset_vector, new_recursive.offset_save,
1642                new_recursive.saved_max * sizeof(int));
1643            md->recursive = new_recursive.prevrec;
1644          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1645            {            {
1646            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1647            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1648              (pcre_free)(new_recursive.offset_save);              (pcre_free)(new_recursive.offset_save);
1649            MRRETURN(MATCH_MATCH);  
1650              /* Set where we got to in the subject, and reset the start in case
1651              it was changed by \K. This *is* propagated back out of a recursion,
1652              for Perl compatibility. */
1653    
1654              eptr = md->end_match_ptr;
1655              mstart = md->start_match_ptr;
1656              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1657            }            }
1658          else if (rrc != MATCH_NOMATCH &&  
1659                  (rrc != MATCH_THEN || md->start_match_ptr != ecode))          /* PCRE does not allow THEN to escape beyond a recursion; it is treated
1660            as NOMATCH. */
1661    
1662            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1663            {            {
1664            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1665            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
# Line 1569  for (;;) Line 1668  for (;;)
1668            }            }
1669    
1670          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1671          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1672          }          }
1673        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1581  for (;;) Line 1678  for (;;)
1678          (pcre_free)(new_recursive.offset_save);          (pcre_free)(new_recursive.offset_save);
1679        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1680        }        }
     /* Control never reaches here */  
   
     /* "Once" brackets are like assertion brackets except that after a match,  
     the point in the subject string is not moved back. Thus there can never be  
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer, but resetting  
     the start-of-match value in case it was changed by \K. */  
1681    
1682      case OP_ONCE:      RECURSION_MATCHED:
1683      prev = ecode;      break;
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);  
       if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */  
         {  
         mstart = md->start_match_ptr;  
         break;  
         }  
       if (rrc != MATCH_NOMATCH &&  
           (rrc != MATCH_THEN || md->start_match_ptr != ecode))  
         RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
   
     /* If hit the end of the group (which could be repeated), fail */  
   
     if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);  
   
     /* Continue as from after the assertion, updating the offsets high water  
     mark, since extracts may have been taken. */  
   
     do ecode += GET(ecode, 1); while (*ecode == OP_ALT);  
   
     offset_top = md->end_offset_top;  
     eptr = md->end_match_ptr;  
   
     /* For a non-repeating ket, just continue at this level. This also  
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
       {  
       ecode += 1+LINK_SIZE;  
       break;  
       }  
   
     /* The repeating kets try the rest of the pattern or restart from the  
     preceding bracket, in the appropriate order. The second "call" of match()  
     uses tail recursion, to avoid using another stack frame. */  
   
     if (*ecode == OP_KETRMIN)  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM8);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       md->match_function_type = MATCH_CBEGROUP;  
       RMATCH(eptr, prev, offset_top, md, eptrb, RM9);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
1684    
1685      /* An alternation is the end of a branch; scan along to find the end of the      /* An alternation is the end of a branch; scan along to find the end of the
1686      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1667  for (;;) Line 1694  for (;;)
1694      at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets      at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
1695      with fixed upper repeat limits are compiled as a number of copies, with the      with fixed upper repeat limits are compiled as a number of copies, with the
1696      optional ones preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1697    
1698      case OP_BRAZERO:      case OP_BRAZERO:
1699      next = ecode + 1;      next = ecode + 1;
1700      RMATCH(eptr, next, offset_top, md, eptrb, RM10);      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
# Line 1675  for (;;) Line 1702  for (;;)
1702      do next += GET(next, 1); while (*next == OP_ALT);      do next += GET(next, 1); while (*next == OP_ALT);
1703      ecode = next + 1 + LINK_SIZE;      ecode = next + 1 + LINK_SIZE;
1704      break;      break;
1705    
1706      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1707      next = ecode + 1;      next = ecode + 1;
1708      do next += GET(next, 1); while (*next == OP_ALT);      do next += GET(next, 1); while (*next == OP_ALT);
# Line 1689  for (;;) Line 1716  for (;;)
1716      do next += GET(next,1); while (*next == OP_ALT);      do next += GET(next,1); while (*next == OP_ALT);
1717      ecode = next + 1 + LINK_SIZE;      ecode = next + 1 + LINK_SIZE;
1718      break;      break;
1719    
1720      /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything      /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1721      here; just jump to the group, with allow_zero set TRUE. */      here; just jump to the group, with allow_zero set TRUE. */
1722    
1723      case OP_BRAPOSZERO:      case OP_BRAPOSZERO:
1724      op = *(++ecode);      op = *(++ecode);
1725      allow_zero = TRUE;      allow_zero = TRUE;
1726      if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;      if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1727        goto POSSESSIVE_NON_CAPTURE;        goto POSSESSIVE_NON_CAPTURE;
# Line 1704  for (;;) Line 1731  for (;;)
1731      case OP_KET:      case OP_KET:
1732      case OP_KETRMIN:      case OP_KETRMIN:
1733      case OP_KETRMAX:      case OP_KETRMAX:
1734      case OP_KETRPOS:      case OP_KETRPOS:
1735      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1736    
1737      /* If this was a group that remembered the subject start, in order to break      /* If this was a group that remembered the subject start, in order to break
1738      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1739      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1740    
1741      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1742        {        {
1743        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1744        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1745        }        }
1746      else saved_eptr = NULL;      else saved_eptr = NULL;
1747    
1748      /* If we are at the end of an assertion group or an atomic group, stop      /* If we are at the end of an assertion group, stop matching and return
1749      matching and return MATCH_MATCH, but record the current high water mark for      MATCH_MATCH, but record the current high water mark for use by positive
1750      use by positive assertions. We also need to record the match start in case      assertions. We also need to record the match start in case it was changed
1751      it was changed by \K. */      by \K. */
1752    
1753      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||      if (*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT)
         *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||  
         *prev == OP_ONCE)  
1754        {        {
1755        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE */
1756        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1757        md->start_match_ptr = mstart;        md->start_match_ptr = mstart;
1758        MRRETURN(MATCH_MATCH);        MRRETURN(MATCH_MATCH);         /* Sets md->mark */
1759        }        }
1760    
1761      /* For capturing groups we have to check the group number back at the start      /* For capturing groups we have to check the group number back at the start
1762      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1763      bumping the high water mark. Note that whole-pattern recursion is coded as      bumping the high water mark. Whole-pattern recursion is coded as a recurse
1764      a recurse into group 0, so it won't be picked up here. Instead, we catch it      into group 0, so it won't be picked up here. Instead, we catch it when the
1765      when the OP_END is reached. Other recursion is handled here. */      OP_END is reached. Other recursion is handled here. We just have to record
1766        the current subject position and start match pointer and give a MATCH
1767        return. */
1768    
1769      if (*prev == OP_CBRA || *prev == OP_SCBRA ||      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1770          *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)          *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
# Line 1750  for (;;) Line 1777  for (;;)
1777        printf("\n");        printf("\n");
1778  #endif  #endif
1779    
1780          /* Handle a recursively called group. */
1781    
1782          if (md->recursive != NULL && md->recursive->group_num == number)
1783            {
1784            md->end_match_ptr = eptr;
1785            md->start_match_ptr = mstart;
1786            RRETURN(MATCH_MATCH);
1787            }
1788    
1789          /* Deal with capturing */
1790    
1791        md->capture_last = number;        md->capture_last = number;
1792        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1793          {          {
1794            /* If offset is greater than offset_top, it means that we are
1795            "skipping" a capturing group, and that group's offsets must be marked
1796            unset. In earlier versions of PCRE, all the offsets were unset at the
1797            start of matching, but this doesn't work because atomic groups and
1798            assertions can cause a value to be set that should later be unset.
1799            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1800            part of the atomic group, but this is not on the final matching path,
1801            so must be unset when 2 is set. (If there is no group 2, there is no
1802            problem, because offset_top will then be 2, indicating no capture.) */
1803    
1804            if (offset > offset_top)
1805              {
1806              register int *iptr = md->offset_vector + offset_top;
1807              register int *iend = md->offset_vector + offset;
1808              while (iptr < iend) *iptr++ = -1;
1809              }
1810    
1811            /* Now make the extraction */
1812    
1813          md->offset_vector[offset] =          md->offset_vector[offset] =
1814            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1815          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1816          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1817          }          }
   
       /* Handle a recursively called group. Restore the offsets  
       appropriately and continue from after the call. */  
   
       if (md->recursive != NULL && md->recursive->group_num == number)  
         {  
         recursion_info *rec = md->recursive;  
         DPRINTF(("Recursion (%d) succeeded - continuing\n", number));  
         md->recursive = rec->prevrec;  
         memcpy(md->offset_vector, rec->offset_save,  
           rec->saved_max * sizeof(int));  
         offset_top = rec->save_offset_top;  
         ecode = rec->after_call;  
         break;  
         }  
1818        }        }
1819    
1820      /* For a non-repeating ket, just continue at this level. This also      /* For an ordinary non-repeating ket, just continue at this level. This
1821      happens for a repeating ket if no characters were matched in the group.      also happens for a repeating ket if no characters were matched in the
1822      This is the forcible breaking of infinite loops as implemented in Perl      group. This is the forcible breaking of infinite loops as implemented in
1823      5.005. If there is an options reset, it will get obeyed in the normal      Perl 5.005. For a non-repeating atomic group, establish a backup point by
1824      course of events. */      processing the rest of the pattern at a lower level. If this results in a
1825        NOMATCH return, pass MATCH_ONCE back to the original OP_ONCE level, thereby
1826        bypassing intermediate backup points, but resetting any captures that
1827        happened along the way. */
1828    
1829      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1830        {        {
1831        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
1832            {
1833            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1834            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1835            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1836            RRETURN(MATCH_ONCE);
1837            }
1838          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1839        break;        break;
1840        }        }
1841    
1842      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1843      and return the MATCH_KETRPOS. This makes it possible to do the repeats one      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1844      at a time from the outer level, thus saving stack. */      at a time from the outer level, thus saving stack. */
1845    
1846      if (*ecode == OP_KETRPOS)      if (*ecode == OP_KETRPOS)
1847        {        {
1848        md->end_match_ptr = eptr;        md->end_match_ptr = eptr;
1849        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1850        RRETURN(MATCH_KETRPOS);        RRETURN(MATCH_KETRPOS);
1851        }        }
1852    
1853      /* The normal repeating kets try the rest of the pattern or restart from      /* The normal repeating kets try the rest of the pattern or restart from
1854      the preceding bracket, in the appropriate order. In the second case, we can      the preceding bracket, in the appropriate order. In the second case, we can
1855      use tail recursion to avoid using another stack frame, unless we have an      use tail recursion to avoid using another stack frame, unless we have an
1856      unlimited repeat of a group that can match an empty string. */      an atomic group or an unlimited repeat of a group that can match an empty
1857        string. */
1858    
1859      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1860        {        {
1861        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1862        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1863          if (*prev == OP_ONCE)
1864            {
1865            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
1866            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1867            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1868            RRETURN(MATCH_ONCE);
1869            }
1870        if (*prev >= OP_SBRA)    /* Could match an empty string */        if (*prev >= OP_SBRA)    /* Could match an empty string */
1871          {          {
1872          md->match_function_type = MATCH_CBEGROUP;          md->match_function_type = MATCH_CBEGROUP;
1873          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1874          RRETURN(rrc);          RRETURN(rrc);
1875          }          }
# Line 1818  for (;;) Line 1878  for (;;)
1878        }        }
1879      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
1880        {        {
1881        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1882        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1883          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
1884        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1885          if (*prev == OP_ONCE)
1886            {
1887            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
1888            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1889            md->once_target = prev;
1890            RRETURN(MATCH_ONCE);
1891            }
1892        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1893        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1894        }        }
# Line 1830  for (;;) Line 1898  for (;;)
1898    
1899      case OP_CIRC:      case OP_CIRC:
1900      if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1901    
1902      /* Start of subject assertion */      /* Start of subject assertion */
1903    
1904      case OP_SOD:      case OP_SOD:
1905      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);
1906      ecode++;      ecode++;
1907      break;      break;
1908    
1909      /* Multiline mode: start of subject unless notbol, or after any newline. */      /* Multiline mode: start of subject unless notbol, or after any newline. */
1910    
1911      case OP_CIRCM:      case OP_CIRCM:
# Line 1876  for (;;) Line 1944  for (;;)
1944      ecode++;      ecode++;
1945      break;      break;
1946    
1947      /* Not multiline mode: assert before a terminating newline or before end of      /* Not multiline mode: assert before a terminating newline or before end of
1948      subject unless noteol is set. */      subject unless noteol is set. */
1949    
1950      case OP_DOLL:      case OP_DOLL:
# Line 2032  for (;;) Line 2100  for (;;)
2100      /* Fall through */      /* Fall through */
2101    
2102      case OP_ALLANY:      case OP_ALLANY:
2103      if (eptr++ >= md->end_subject)      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2104        {        {                            /* not be updated before SCHECK_PARTIAL. */
2105        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2106        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2107        }        }
2108        eptr++;
2109      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
2110      ecode++;      ecode++;
2111      break;      break;
# Line 2045  for (;;) Line 2114  for (;;)
2114      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2115    
2116      case OP_ANYBYTE:      case OP_ANYBYTE:
2117      if (eptr++ >= md->end_subject)      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2118        {        {                            /* not be updated before SCHECK_PARTIAL. */
2119        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2120        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2121        }        }
2122        eptr++;
2123      ecode++;      ecode++;
2124      break;      break;
2125    
# Line 2165  for (;;) Line 2235  for (;;)
2235      switch(c)      switch(c)
2236        {        {
2237        default: MRRETURN(MATCH_NOMATCH);        default: MRRETURN(MATCH_NOMATCH);
2238    
2239        case 0x000d:        case 0x000d:
2240        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2241        break;        break;
# Line 2389  for (;;) Line 2459  for (;;)
2459        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2460        }        }
2461      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2462        if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
2463        while (eptr < md->end_subject)
2464        {        {
2465        int category = UCD_CATEGORY(c);        int len = 1;
2466        if (category == ucp_M) MRRETURN(MATCH_NOMATCH);        if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2467        while (eptr < md->end_subject)        if (UCD_CATEGORY(c) != ucp_M) break;
2468          {        eptr += len;
         int len = 1;  
         if (!utf8) c = *eptr; else  
           {  
           GETCHARLEN(c, eptr, len);  
           }  
         category = UCD_CATEGORY(c);  
         if (category != ucp_M) break;  
         eptr += len;  
         }  
2469        }        }
2470      ecode++;      ecode++;
2471      break;      break;
# Line 2418  for (;;) Line 2481  for (;;)
2481      loops). */      loops). */
2482    
2483      case OP_REF:      case OP_REF:
2484      case OP_REFI:      case OP_REFI:
2485      caseless = op == OP_REFI;      caseless = op == OP_REFI;
2486      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2487      ecode += 3;      ecode += 3;
2488    
# Line 2487  for (;;) Line 2550  for (;;)
2550    
2551      for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2552        {        {
2553        int slength;        int slength;
2554        if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)        if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2555          {          {
2556          CHECK_PARTIAL();          CHECK_PARTIAL();
# Line 2507  for (;;) Line 2570  for (;;)
2570        {        {
2571        for (fi = min;; fi++)        for (fi = min;; fi++)
2572          {          {
2573          int slength;          int slength;
2574          RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);          RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2575          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2576          if (fi >= max) MRRETURN(MATCH_NOMATCH);          if (fi >= max) MRRETURN(MATCH_NOMATCH);
# Line 2528  for (;;) Line 2591  for (;;)
2591        pp = eptr;        pp = eptr;
2592        for (i = min; i < max; i++)        for (i = min; i < max; i++)
2593          {          {
2594          int slength;          int slength;
2595          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2596            {            {
2597            CHECK_PARTIAL();            CHECK_PARTIAL();
# Line 3261  for (;;) Line 3324  for (;;)
3324      checking can be multibyte. */      checking can be multibyte. */
3325    
3326      case OP_NOT:      case OP_NOT:
3327      case OP_NOTI:      case OP_NOTI:
3328      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
3329        {        {
3330        SCHECK_PARTIAL();        SCHECK_PARTIAL();
# Line 3756  for (;;) Line 3819  for (;;)
3819            case PT_LAMP:            case PT_LAMP:
3820            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3821              {              {
3822                int chartype;
3823              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3824                {                {
3825                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3826                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3827                }                }
3828              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3829              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
3830              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
3831                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
3832                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
3833                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3834              }              }
3835            break;            break;
# Line 3779  for (;;) Line 3843  for (;;)
3843                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3844                }                }
3845              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3846              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
             if ((prop_category == prop_value) == prop_fail_result)  
3847                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3848              }              }
3849            break;            break;
# Line 3794  for (;;) Line 3857  for (;;)
3857                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3858                }                }
3859              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3860              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
             if ((prop_chartype == prop_value) == prop_fail_result)  
3861                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3862              }              }
3863            break;            break;
# Line 3809  for (;;) Line 3871  for (;;)
3871                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3872                }                }
3873              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3874              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
             if ((prop_script == prop_value) == prop_fail_result)  
3875                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3876              }              }
3877            break;            break;
# Line 3818  for (;;) Line 3879  for (;;)
3879            case PT_ALNUM:            case PT_ALNUM:
3880            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3881              {              {
3882                int category;
3883              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3884                {                {
3885                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3886                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3887                }                }
3888              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3889              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
3890              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                    == prop_fail_result)  
3891                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3892              }              }
3893            break;            break;
# Line 3840  for (;;) Line 3901  for (;;)
3901                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3902                }                }
3903              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3904              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
3905                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
3906                     == prop_fail_result)                     == prop_fail_result)
3907                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 3857  for (;;) Line 3917  for (;;)
3917                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3918                }                }
3919              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3920              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
3921                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
3922                     == prop_fail_result)                     == prop_fail_result)
3923                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 3868  for (;;) Line 3927  for (;;)
3927            case PT_WORD:            case PT_WORD:
3928            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3929              {              {
3930                int category;
3931              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3932                {                {
3933                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3934                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3935                }                }
3936              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3937              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
3938              if ((prop_category == ucp_L || prop_category == ucp_N ||              if ((category == ucp_L || category == ucp_N || c == CHAR_UNDERSCORE)
                  c == CHAR_UNDERSCORE)  
3939                     == prop_fail_result)                     == prop_fail_result)
3940                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3941              }              }
# Line 3902  for (;;) Line 3961  for (;;)
3961              MRRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
3962              }              }
3963            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
3964            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
           if (prop_category == ucp_M) MRRETURN(MATCH_NOMATCH);  
3965            while (eptr < md->end_subject)            while (eptr < md->end_subject)
3966              {              {
3967              int len = 1;              int len = 1;
3968              if (!utf8) c = *eptr;              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
3969                else { GETCHARLEN(c, eptr, len); }              if (UCD_CATEGORY(c) != ucp_M) break;
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
3970              eptr += len;              eptr += len;
3971              }              }
3972            }            }
# Line 3968  for (;;) Line 4024  for (;;)
4024            switch(c)            switch(c)
4025              {              {
4026              default: MRRETURN(MATCH_NOMATCH);              default: MRRETURN(MATCH_NOMATCH);
4027    
4028              case 0x000d:              case 0x000d:
4029              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
4030              break;              break;
# Line 4245  for (;;) Line 4301  for (;;)
4301            switch(*eptr++)            switch(*eptr++)
4302              {              {
4303              default: MRRETURN(MATCH_NOMATCH);              default: MRRETURN(MATCH_NOMATCH);
4304    
4305              case 0x000d:              case 0x000d:
4306              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
4307              break;              break;
4308    
4309              case 0x000a:              case 0x000a:
4310              break;              break;
4311    
# Line 4455  for (;;) Line 4511  for (;;)
4511            case PT_LAMP:            case PT_LAMP:
4512            for (fi = min;; fi++)            for (fi = min;; fi++)
4513              {              {
4514                int chartype;
4515              RMATCH(eptr, ecode, offset_top, md, eptrb, RM37);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM37);
4516              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4517              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
# Line 4464  for (;;) Line 4521  for (;;)
4521                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4522                }                }
4523              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4524              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
4525              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
4526                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
4527                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
4528                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4529              }              }
4530            /* Control never gets here */            /* Control never gets here */
# Line 4484  for (;;) Line 4541  for (;;)
4541                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4542                }                }
4543              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4544              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
             if ((prop_category == prop_value) == prop_fail_result)  
4545                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4546              }              }
4547            /* Control never gets here */            /* Control never gets here */
# Line 4502  for (;;) Line 4558  for (;;)
4558                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4559                }                }
4560              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4561              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
             if ((prop_chartype == prop_value) == prop_fail_result)  
4562                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4563              }              }
4564            /* Control never gets here */            /* Control never gets here */
# Line 4520  for (;;) Line 4575  for (;;)
4575                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4576                }                }
4577              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4578              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
             if ((prop_script == prop_value) == prop_fail_result)  
4579                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4580              }              }
4581            /* Control never gets here */            /* Control never gets here */
# Line 4529  for (;;) Line 4583  for (;;)
4583            case PT_ALNUM:            case PT_ALNUM:
4584            for (fi = min;; fi++)            for (fi = min;; fi++)
4585              {              {
4586                int category;
4587              RMATCH(eptr, ecode, offset_top, md, eptrb, RM59);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM59);
4588              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4589              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
# Line 4538  for (;;) Line 4593  for (;;)
4593                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4594                }                }
4595              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4596              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
4597              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                    == prop_fail_result)  
4598                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4599              }              }
4600            /* Control never gets here */            /* Control never gets here */
# Line 4557  for (;;) Line 4611  for (;;)
4611                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4612                }                }
4613              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4614              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
4615                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
4616                     == prop_fail_result)                     == prop_fail_result)
4617                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4577  for (;;) Line 4630  for (;;)
4630                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4631                }                }
4632              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4633              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
4634                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
4635                     == prop_fail_result)                     == prop_fail_result)
4636                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4588  for (;;) Line 4640  for (;;)
4640            case PT_WORD:            case PT_WORD:
4641            for (fi = min;; fi++)            for (fi = min;; fi++)
4642              {              {
4643                int category;
4644              RMATCH(eptr, ecode, offset_top, md, eptrb, RM62);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM62);
4645              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4646              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
# Line 4597  for (;;) Line 4650  for (;;)
4650                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4651                }                }
4652              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4653              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
4654              if ((prop_category == ucp_L ||              if ((category == ucp_L ||
4655                   prop_category == ucp_N ||                   category == ucp_N ||
4656                   c == CHAR_UNDERSCORE)                   c == CHAR_UNDERSCORE)
4657                     == prop_fail_result)                     == prop_fail_result)
4658                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4629  for (;;) Line 4682  for (;;)
4682              MRRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
4683              }              }
4684            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
4685            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
           if (prop_category == ucp_M) MRRETURN(MATCH_NOMATCH);  
4686            while (eptr < md->end_subject)            while (eptr < md->end_subject)
4687              {              {
4688              int len = 1;              int len = 1;
4689              if (!utf8) c = *eptr;              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
4690                else { GETCHARLEN(c, eptr, len); }              if (UCD_CATEGORY(c) != ucp_M) break;
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
4691              eptr += len;              eptr += len;
4692              }              }
4693            }            }
4694          }          }
   
4695        else        else
4696  #endif     /* SUPPORT_UCP */  #endif     /* SUPPORT_UCP */
4697    
# Line 4963  for (;;) Line 5012  for (;;)
5012            case PT_LAMP:            case PT_LAMP:
5013            for (i = min; i < max; i++)            for (i = min; i < max; i++)
5014              {              {
5015                int chartype;
5016              int len = 1;              int len = 1;
5017              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
5018                {                {
# Line 4970  for (;;) Line 5020  for (;;)
5020                break;                break;
5021                }                }
5022              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5023              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
5024              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
5025                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
5026                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
5027                break;                break;
5028              eptr+= len;              eptr+= len;
5029              }              }
# Line 4989  for (;;) Line 5039  for (;;)
5039                break;                break;
5040                }                }
5041              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5042              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) break;
             if ((prop_category == prop_value) == prop_fail_result)  
               break;  
5043              eptr+= len;              eptr+= len;
5044              }              }
5045            break;            break;
# Line 5006  for (;;) Line 5054  for (;;)
5054                break;                break;
5055                }                }
5056              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5057              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) break;
             if ((prop_chartype == prop_value) == prop_fail_result)  
               break;  
5058              eptr+= len;              eptr+= len;
5059              }              }
5060            break;            break;
# Line 5023  for (;;) Line 5069  for (;;)
5069                break;                break;
5070                }                }
5071              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5072              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) break;
             if ((prop_script == prop_value) == prop_fail_result)  
               break;  
5073              eptr+= len;              eptr+= len;
5074              }              }
5075            break;            break;
# Line 5033  for (;;) Line 5077  for (;;)
5077            case PT_ALNUM:            case PT_ALNUM:
5078            for (i = min; i < max; i++)            for (i = min; i < max; i++)
5079              {              {
5080                int category;
5081              int len = 1;              int len = 1;
5082              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
5083                {                {
# Line 5040  for (;;) Line 5085  for (;;)
5085                break;                break;
5086                }                }
5087              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5088              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
5089              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                  == prop_fail_result)  
5090                break;                break;
5091              eptr+= len;              eptr+= len;
5092              }              }
# Line 5058  for (;;) Line 5102  for (;;)
5102                break;                break;
5103                }                }
5104              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5105              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
5106                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
5107                   == prop_fail_result)                   == prop_fail_result)
5108                break;                break;
# Line 5077  for (;;) Line 5120  for (;;)
5120                break;                break;
5121                }                }
5122              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5123              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
5124                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
5125                   == prop_fail_result)                   == prop_fail_result)
5126                break;                break;
# Line 5089  for (;;) Line 5131  for (;;)
5131            case PT_WORD:            case PT_WORD:
5132            for (i = min; i < max; i++)            for (i = min; i < max; i++)
5133              {              {
5134                int category;
5135              int len = 1;              int len = 1;
5136              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
5137                {                {
# Line 5096  for (;;) Line 5139  for (;;)
5139                break;                break;
5140                }                }
5141              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5142              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
5143              if ((prop_category == ucp_L || prop_category == ucp_N ||              if ((category == ucp_L || category == ucp_N ||
5144                   c == CHAR_UNDERSCORE) == prop_fail_result)                   c == CHAR_UNDERSCORE) == prop_fail_result)
5145                break;                break;
5146              eptr+= len;              eptr+= len;
# Line 5127  for (;;) Line 5170  for (;;)
5170          {          {
5171          for (i = min; i < max; i++)          for (i = min; i < max; i++)
5172            {            {
5173              int len = 1;
5174            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
5175              {              {
5176              SCHECK_PARTIAL();              SCHECK_PARTIAL();
5177              break;              break;
5178              }              }
5179            GETCHARINCTEST(c, eptr);            if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5180            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) break;
5181            if (prop_category == ucp_M) break;            eptr += len;
5182            while (eptr < md->end_subject)            while (eptr < md->end_subject)
5183              {              {
5184              int len = 1;              len = 1;
5185              if (!utf8) c = *eptr; else              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5186                {              if (UCD_CATEGORY(c) != ucp_M) break;
               GETCHARLEN(c, eptr, len);  
               }  
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
5187              eptr += len;              eptr += len;
5188              }              }
5189            }            }
# Line 5159  for (;;) Line 5199  for (;;)
5199            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
5200            for (;;)                        /* Move back over one extended */            for (;;)                        /* Move back over one extended */
5201              {              {
             int len = 1;  
5202              if (!utf8) c = *eptr; else              if (!utf8) c = *eptr; else
5203                {                {
5204                BACKCHAR(eptr);                BACKCHAR(eptr);
5205                GETCHARLEN(c, eptr, len);                GETCHAR(c, eptr);
5206                }                }
5207              prop_category = UCD_CATEGORY(c);              if (UCD_CATEGORY(c) != ucp_M) break;
             if (prop_category != ucp_M) break;  
5208              eptr--;              eptr--;
5209              }              }
5210            }            }
# Line 5230  for (;;) Line 5268  for (;;)
5268                while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;                while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
5269                }                }
5270              }              }
5271            else eptr = md->end_subject;   /* Unlimited UTF-8 repeat */            else
5272                {
5273                eptr = md->end_subject;   /* Unlimited UTF-8 repeat */
5274                SCHECK_PARTIAL();
5275                }
5276            break;            break;
5277    
5278            /* The byte case is the same as non-UTF8 */            /* The byte case is the same as non-UTF8 */
# Line 5451  for (;;) Line 5493  for (;;)
5493            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5494            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
5495            BACKCHAR(eptr);            BACKCHAR(eptr);
5496            if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&            if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&
5497                eptr[-1] == '\r') eptr--;                eptr[-1] == '\r') eptr--;
5498            }            }
5499          }          }
# Line 5663  for (;;) Line 5705  for (;;)
5705            RMATCH(eptr, ecode, offset_top, md, eptrb, RM47);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM47);
5706            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5707            eptr--;            eptr--;
5708            if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&            if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&
5709                eptr[-1] == '\r') eptr--;                eptr[-1] == '\r') eptr--;
5710            }            }
5711          }          }
# Line 5703  switch (frame->Xwhere) Line 5745  switch (frame->Xwhere)
5745    LBL( 9) LBL(10) LBL(11) LBL(12) LBL(13) LBL(14) LBL(15) LBL(17)    LBL( 9) LBL(10) LBL(11) LBL(12) LBL(13) LBL(14) LBL(15) LBL(17)
5746    LBL(19) LBL(24) LBL(25) LBL(26) LBL(27) LBL(29) LBL(31) LBL(33)    LBL(19) LBL(24) LBL(25) LBL(26) LBL(27) LBL(29) LBL(31) LBL(33)
5747    LBL(35) LBL(43) LBL(47) LBL(48) LBL(49) LBL(50) LBL(51) LBL(52)    LBL(35) LBL(43) LBL(47) LBL(48) LBL(49) LBL(50) LBL(51) LBL(52)
5748    LBL(53) LBL(54) LBL(55) LBL(56) LBL(57) LBL(58) LBL(63) LBL(64)    LBL(53) LBL(54) LBL(55) LBL(56) LBL(57) LBL(58) LBL(63)
5749  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
5750    LBL(16) LBL(18) LBL(20) LBL(21) LBL(22) LBL(23) LBL(28) LBL(30)    LBL(16) LBL(18) LBL(20) LBL(21) LBL(22) LBL(23) LBL(28) LBL(30)
5751    LBL(32) LBL(34) LBL(42) LBL(46)    LBL(32) LBL(34) LBL(42) LBL(46)
# Line 5805  pcre_exec(const pcre *argument_re, const Line 5847  pcre_exec(const pcre *argument_re, const
5847    PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,    PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,
5848    int offsetcount)    int offsetcount)
5849  {  {
5850  int rc, resetcount, ocount;  int rc, ocount, arg_offset_max;
5851  int first_byte = -1;  int first_byte = -1;
5852  int req_byte = -1;  int req_byte = -1;
5853  int req_byte2 = -1;  int req_byte2 = -1;
# Line 5841  if (re == NULL || subject == NULL || Line 5883  if (re == NULL || subject == NULL ||
5883  if (offsetcount < 0) return PCRE_ERROR_BADCOUNT;  if (offsetcount < 0) return PCRE_ERROR_BADCOUNT;
5884  if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET;  if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET;
5885    
5886  /* This information is for finding all the numbers associated with a given  /* These two settings are used in the code for checking a UTF-8 string that
5887  name, for condition testing. */  follows immediately afterwards. Other values in the md block are used only
5888    during "normal" pcre_exec() processing, not when the JIT support is in use,
5889    so they are set up later. */
5890    
5891    utf8 = md->utf8 = (re->options & PCRE_UTF8) != 0;
5892    md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :
5893                  ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;
5894    
5895    /* Check a UTF-8 string if required. Pass back the character offset and error
5896    code for an invalid string if a results vector is available. */
5897    
5898    #ifdef SUPPORT_UTF8
5899    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)
5900      {
5901      int erroroffset;
5902      int errorcode = _pcre_valid_utf8((USPTR)subject, length, &erroroffset);
5903      if (errorcode != 0)
5904        {
5905        if (offsetcount >= 2)
5906          {
5907          offsets[0] = erroroffset;
5908          offsets[1] = errorcode;
5909          }
5910        return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)?
5911          PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;
5912        }
5913    
5914      /* Check that a start_offset points to the start of a UTF-8 character. */
5915      if (start_offset > 0 && start_offset < length &&
5916          (((USPTR)subject)[start_offset] & 0xc0) == 0x80)
5917        return PCRE_ERROR_BADUTF8_OFFSET;
5918      }
5919    #endif
5920    
5921    /* If the pattern was successfully studied with JIT support, run the JIT
5922    executable instead of the rest of this function. Most options must be set at
5923    compile time for the JIT code to be usable. Fallback to the normal code path if
5924    an unsupported flag is set. In particular, JIT does not support partial
5925    matching. */
5926    
5927    #ifdef SUPPORT_JIT
5928    if (extra_data != NULL
5929        && (extra_data->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0
5930        && extra_data->executable_jit != NULL
5931        && (options & ~(PCRE_NO_UTF8_CHECK | PCRE_NOTBOL | PCRE_NOTEOL |
5932                        PCRE_NOTEMPTY | PCRE_NOTEMPTY_ATSTART)) == 0)
5933      return _pcre_jit_exec(re, extra_data->executable_jit, subject, length,
5934        start_offset, options, ((extra_data->flags & PCRE_EXTRA_MATCH_LIMIT) == 0)
5935        ? MATCH_LIMIT : extra_data->match_limit, offsets, offsetcount);
5936    #endif
5937    
5938    /* Carry on with non-JIT matching. This information is for finding all the
5939    numbers associated with a given name, for condition testing. */
5940    
5941  md->name_table = (uschar *)re + re->name_table_offset;  md->name_table = (uschar *)re + re->name_table_offset;
5942  md->name_count = re->name_count;  md->name_count = re->name_count;
# Line 5909  md->end_subject = md->start_subject + le Line 6003  md->end_subject = md->start_subject + le
6003  end_subject = md->end_subject;  end_subject = md->end_subject;
6004    
6005  md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;  md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
 utf8 = md->utf8 = (re->options & PCRE_UTF8) != 0;  
6006  md->use_ucp = (re->options & PCRE_UCP) != 0;  md->use_ucp = (re->options & PCRE_UCP) != 0;
6007  md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;  md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;
6008    
6009    /* Some options are unpacked into BOOL variables in the hope that testing
6010    them will be faster than individual option bits. */
6011    
6012  md->notbol = (options & PCRE_NOTBOL) != 0;  md->notbol = (options & PCRE_NOTBOL) != 0;
6013  md->noteol = (options & PCRE_NOTEOL) != 0;  md->noteol = (options & PCRE_NOTEOL) != 0;
6014  md->notempty = (options & PCRE_NOTEMPTY) != 0;  md->notempty = (options & PCRE_NOTEMPTY) != 0;
6015  md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;  md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;
6016  md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :  
               ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;  
6017  md->hitend = FALSE;  md->hitend = FALSE;
6018  md->mark = NULL;                        /* In case never set */  md->mark = NULL;                        /* In case never set */
6019    
6020  md->recursive = NULL;                   /* No recursion at top level */  md->recursive = NULL;                   /* No recursion at top level */
6021    md->hasthen = (re->flags & PCRE_HASTHEN) != 0;
6022    
6023  md->lcc = tables + lcc_offset;  md->lcc = tables + lcc_offset;
6024  md->ctypes = tables + ctypes_offset;  md->ctypes = tables + ctypes_offset;
# Line 6000  defined (though never set). So there's n Line 6096  defined (though never set). So there's n
6096  if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)  if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)
6097    return PCRE_ERROR_BADPARTIAL;    return PCRE_ERROR_BADPARTIAL;
6098    
 /* Check a UTF-8 string if required. Pass back the character offset and error  
 code for an invalid string if a results vector is available. */  
   
 #ifdef SUPPORT_UTF8  
 if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)  
   {  
   int erroroffset;  
   int errorcode = _pcre_valid_utf8((USPTR)subject, length, &erroroffset);  
   if (errorcode != 0)  
     {  
     if (offsetcount >= 2)  
       {  
       offsets[0] = erroroffset;  
       offsets[1] = errorcode;  
       }  
     return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)?  
       PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;  
     }  
   
   /* Check that a start_offset points to the start of a UTF-8 character. */  
   
   if (start_offset > 0 && start_offset < length &&  
       (((USPTR)subject)[start_offset] & 0xc0) == 0x80)  
     return PCRE_ERROR_BADUTF8_OFFSET;  
   }  
 #endif  
   
6099  /* If the expression has got more back references than the offsets supplied can  /* If the expression has got more back references than the offsets supplied can
6100  hold, we get a temporary chunk of working store to use during the matching.  hold, we get a temporary chunk of working store to use during the matching.
6101  Otherwise, we can use the vector supplied, rounding down its size to a multiple  Otherwise, we can use the vector supplied, rounding down its size to a multiple
6102  of 3. */  of 3. */
6103    
6104  ocount = offsetcount - (offsetcount % 3);  ocount = offsetcount - (offsetcount % 3);
6105    arg_offset_max = (2*ocount)/3;
6106    
6107  if (re->top_backref > 0 && re->top_backref >= ocount/3)  if (re->top_backref > 0 && re->top_backref >= ocount/3)
6108    {    {
# Line 6049  md->offset_max = (2*ocount)/3; Line 6119  md->offset_max = (2*ocount)/3;
6119  md->offset_overflow = FALSE;  md->offset_overflow = FALSE;
6120  md->capture_last = -1;  md->capture_last = -1;
6121    
 /* Compute the minimum number of offsets that we need to reset each time. Doing  
 this makes a huge difference to execution time when there aren't many brackets  
 in the pattern. */  
   
 resetcount = 2 + re->top_bracket * 2;  
 if (resetcount > offsetcount) resetcount = ocount;  
   
6122  /* Reset the working variable associated with each extraction. These should  /* Reset the working variable associated with each extraction. These should
6123  never be used unless previously set, but they get saved and restored, and so we  never be used unless previously set, but they get saved and restored, and so we
6124  initialize them to avoid reading uninitialized locations. */  initialize them to avoid reading uninitialized locations. Also, unset the
6125    offsets for the matched string. This is really just for tidiness with callouts,
6126    in case they inspect these fields. */
6127    
6128  if (md->offset_vector != NULL)  if (md->offset_vector != NULL)
6129    {    {
6130    register int *iptr = md->offset_vector + ocount;    register int *iptr = md->offset_vector + ocount;
6131    register int *iend = iptr - resetcount/2 + 1;    register int *iend = iptr - re->top_bracket;
6132      if (iend < md->offset_vector + 2) iend = md->offset_vector + 2;
6133    while (--iptr >= iend) *iptr = -1;    while (--iptr >= iend) *iptr = -1;
6134      md->offset_vector[0] = md->offset_vector[1] = -1;
6135    }    }
6136    
6137  /* Set up the first character to match, if available. The first_byte value is  /* Set up the first character to match, if available. The first_byte value is
# Line 6098  if ((re->flags & PCRE_REQCHSET) != 0) Line 6165  if ((re->flags & PCRE_REQCHSET) != 0)
6165    }    }
6166    
6167    
6168    
6169    
6170  /* ==========================================================================*/  /* ==========================================================================*/
6171    
6172  /* Loop for handling unanchored repeated matching attempts; for anchored regexs  /* Loop for handling unanchored repeated matching attempts; for anchored regexs
# Line 6108  for(;;) Line 6177  for(;;)
6177    USPTR save_end_subject = end_subject;    USPTR save_end_subject = end_subject;
6178    USPTR new_start_match;    USPTR new_start_match;
6179    
   /* Reset the maximum number of extractions we might see. */  
   
   if (md->offset_vector != NULL)  
     {  
     register int *iptr = md->offset_vector;  
     register int *iend = iptr + resetcount;  
     while (iptr < iend) *iptr++ = -1;  
     }  
   
6180    /* If firstline is TRUE, the start of the match is constrained to the first    /* If firstline is TRUE, the start of the match is constrained to the first
6181    line of a multiline string. That is, the match must be before or at the first    line of a multiline string. That is, the match must be before or at the first
6182    newline. Implement this by temporarily adjusting end_subject so that we stop    newline. Implement this by temporarily adjusting end_subject so that we stop
# Line 6305  for(;;) Line 6365  for(;;)
6365    md->start_match_ptr = start_match;    md->start_match_ptr = start_match;
6366    md->start_used_ptr = start_match;    md->start_used_ptr = start_match;
6367    md->match_call_count = 0;    md->match_call_count = 0;
6368    md->match_function_type = 0;    md->match_function_type = 0;
6369      md->end_offset_top = 0;
6370    rc = match(start_match, md->start_code, start_match, NULL, 2, md, NULL, 0);    rc = match(start_match, md->start_code, start_match, NULL, 2, md, NULL, 0);
6371    if (md->hitend && start_partial == NULL) start_partial = md->start_used_ptr;    if (md->hitend && start_partial == NULL) start_partial = md->start_used_ptr;
6372    
# Line 6416  if (rc == MATCH_MATCH || rc == MATCH_ACC Line 6477  if (rc == MATCH_MATCH || rc == MATCH_ACC
6477    {    {
6478    if (using_temporary_offsets)    if (using_temporary_offsets)
6479      {      {
6480      if (offsetcount >= 4)      if (arg_offset_max >= 4)
6481        {        {
6482        memcpy(offsets + 2, md->offset_vector + 2,        memcpy(offsets + 2, md->offset_vector + 2,
6483          (offsetcount - 2) * sizeof(int));          (arg_offset_max - 2) * sizeof(int));
6484        DPRINTF(("Copied offsets from temporary memory\n"));        DPRINTF(("Copied offsets from temporary memory\n"));
6485        }        }
6486      if (md->end_offset_top > offsetcount) md->offset_overflow = TRUE;      if (md->end_offset_top > arg_offset_max) md->offset_overflow = TRUE;
6487      DPRINTF(("Freeing temporary memory\n"));      DPRINTF(("Freeing temporary memory\n"));
6488      (pcre_free)(md->offset_vector);      (pcre_free)(md->offset_vector);
6489      }      }
6490    
6491    /* Set the return code to the number of captured strings, or 0 if there are    /* Set the return code to the number of captured strings, or 0 if there were
6492    too many to fit into the vector. */    too many to fit into the vector. */
6493    
6494    rc = md->offset_overflow? 0 : md->end_offset_top/2;    rc = (md->offset_overflow && md->end_offset_top >= arg_offset_max)?
6495        0 : md->end_offset_top/2;
6496    
6497      /* If there is space in the offset vector, set any unused pairs at the end of
6498      the pattern to -1 for backwards compatibility. It is documented that this
6499      happens. In earlier versions, the whole set of potential capturing offsets
6500      was set to -1 each time round the loop, but this is handled differently now.
6501      "Gaps" are set to -1 dynamically instead (this fixes a bug). Thus, it is only
6502      those at the end that need unsetting here. We can't just unset them all at
6503      the start of the whole thing because they may get set in one branch that is
6504      not the final matching branch. */
6505    
6506      if (md->end_offset_top/2 <= re->top_bracket && offsets != NULL)
6507        {
6508        register int *iptr, *iend;
6509        int resetcount = 2 + re->top_bracket * 2;
6510        if (resetcount > offsetcount) resetcount = ocount;
6511        iptr = offsets + md->end_offset_top;
6512        iend = offsets + resetcount;
6513        while (iptr < iend) *iptr++ = -1;
6514        }
6515    
6516    /* If there is space, set up the whole thing as substring 0. The value of    /* If there is space, set up the whole thing as substring 0. The value of
6517    md->start_match_ptr might be modified if \K was encountered on the success    md->start_match_ptr might be modified if \K was encountered on the success

Legend:
Removed from v.608  
changed lines
  Added in v.721

  ViewVC Help
Powered by ViewVC 1.1.5